Summary

In this chapter, we explored the concept of time series, examined the core characteristics and types of time series, and looked at some well-known applications of time series in machine learning. We also covered concepts such as trailing and centered windows and examined how to prepare time series for modeling with neural networks with the aid of utilities from TensorFlow. In our case study, we applied both statistical and deep learning techniques in order to build a sales forecasting model for a fictional company.

In the next chapter, we will extend our modeling using more complex architectures such as RNNs, CNNs, and CNN-LSTM architecture in forecasting time series data. Also, we will explore concepts such as learning rate scheduler and Lambda layers. To conclude the final chapter of this book, we will build a forecasting model for Apple’s closing stock price.